Connecting to a live agent

Important: Connecting to a live agent is currently not supported in on-premises deployment.

As an admin, you can control when and how Orchestrate Chat escalates to human support to ensure timely and accurate assistance. If the chatbot cannot resolve the issue, you can connect users to live agents(human support) directly from the Orchestrate Chat.

Configure the system to automatically connect users to live agents when automation cannot resolve their issue. It involves two primary scenarios:

Live agent integration

Important: Live agent integration is supported only through watsonx Assistant. Other integration methods are not currently available.

Live agent integration allows the chat system to:

  • Transfer chat context to agents
  • Route users to the right support queue

To integrate a live agent, you must:

  1. Create a custom action in your assistant.

    Within the custom action, set up the Connect to Agent logic. It allows the assistant to either transfer the conversation to a human agent or delegate it to another bot, depending on your configuration.

  2. Add that assistant as a collaborator agent to your main agent.

    It allows your main agent to call the assistant when needed. When the assistant is triggered, it can run the custom action and do the handoff.

To enable human support, you need to integrate Orchestrate Chat with a custom service desk.

Adding a custom service desk

This interface allows you to create a custom integration with a service desk or contact center that users can use to communicate with human agents. For more details, see Adding a custom service desk.

Escalation strategy

Escalation paths can be implemented in two ways:

Planned escalation

Planned escalation can be implemented when you anticipate the need to transfer the conversation to a live agent.

Use planned escalation for handling sensitive topics such as:

  • Security concerns (Example: account breaches, unauthorized access)
  • Privacy-related inquiries (Example: data deletion requests, GDPR compliance)
  • Legal or compliance issues (Example: disputes, regulatory questions)
  • Financial matters (Example: billing errors, refund requests)
  • Health or safety-related concerns

Fallback escalation

Fallback escalation can be implemented when the agent cannot resolve the user's issue. Use fallback escalation for situations like:

  • Repeated “I don’t understand” responses
  • User explicitly asking to speak with a human like typing “I want to talk to a human.”
  • Low confidence score in NLP model